Previous   |    Next

doi:10.2204/iodp.proc.314315316.112.2009

Log-seismic correlation

During Expedition 314 we used several of the LWD data sets to establish accurate ties to the 2006 Kumano 3-D seismic reflection data set (Moore et al., 2007). Data from seismicVISION and sonicVISION tools helped to establish a traveltime to depth transform at the boreholes.

seismicVISION tool

The seismicVISION tool produces data which can be interpreted as a check shot survey, a low-resolution velocity depth function, and a vertical seismic profile. The seismicVISION tool records seismograms using a hydrophone and a three-component geophone in the tool and a surface source and hydrophone (www.slb.com/​content/​services/​drilling/​imaging/​seismicvision.asp). The source was three 250 in³ air guns (Fig.F11) that were suspended from crane 1 ~55 m horizontally from the rotary table (Fig.F12) and fired 6 m below sea level at 1700–2000 psi. Time correlations of the shots are ensured using high-precision clocks at both the surface hydrophone and downhole hydrophone. The surface hydrophone was suspended 3 m below the air guns (total 9 m below mean sea level) and the zero times of the waveforms were corrected to mean sea level.

Data cannot be usefully obtained when the pipe is rotating or moving vertically or when the drilling fluid pumps are running. Therefore, seismicVISION data are typically acquired at each addition or removal of a stand of drill pipe (~38 m long) to or from the drill string. The Chikyu uses four joints (9.5 m long drill pipe) as one stand. The drill string is stationary and the pumps are off during these times. This means that while drilling the hole, data were acquired every 38 m. After drilling was complete and while the pipe was being recovered, data were also acquired every 38 m but at points shifted by two joints (19 m) relative to the acquisition during drilling. Thus, combining the data acquired during drilling and the data acquired during pipe recovery yielded data nominally acquired at 19 m intervals in the borehole (Fig. F13).

At each data acquisition level, a number of shots were fired by the surface source. When rig circulation/​rotation stopped, the seismicVISION acquisition system was activated. The source was fired at 15 s intervals. The time for the insertion or removal of a pipe stand was at least 3 min. Thus, the standard procedure was to fire 10–15 shots. The tool records the pressure (hydrophone) and acceleration (three-component geophones) seismograms obtained for each shot and calculates and records a vertical stack of the shots. In practice, because our holes were near vertical and the tool was centered and unclamped, the geophone data yielded little useful information. The tool also automatically picks and records the P-wave arrival time in the stacked seismogram. The stacked seismogram and the picked P-wave arrival times from the instrument are transmitted to the ship by the MWD system and thus could be used soon after acquisition at each level or as the primary data in case the tool fails to record data or is lost.

A check shot survey consists of the one-way first arrival traveltimes of a seismic pulse from the surface to a downhole hydrophone placed at a series of known depths in the hole. This corresponds to the picked P-wave arrival at each depth from the seismicVISION tool. This was the most reliable data set for matching seismic reflectors acquired in the time domain with logging and core data acquired in the depth domain. The check shot data are a key input to the process of making a synthetic seismogram.

We processed the seismicVISION data using ProMAX software (Landmark-Halliburton). After manual editing of bad or noisy traces and frequency filtering, the shots from each level were stacked and first breaks were picked.

A velocity depth function was obtained by taking the differences between adjacent vertical traveltime data (the check shot data) and dividing by the depth difference. Velocity depth curves obtained in this manner are typically quite noisy. A smooth version of this curve was obtained using the methodology of Lizarralde and Swift (1999). This curve may be usefully compared with the velocity versus depth data from the sonicVISION tool and with the velocity versus depth at the hole location used in the processing of the 3-D seismic volume.

A vertical seismic profile can be obtained by assembling a gather of the stacked seismograms from a hole sorted by receiver depth. This gather was filtered to remove noise and the upgoing and downgoing waves were separated using FK filtering. The upgoing arrivals were then moved out (flattened in time) and stacked to produce an equivalent of the zero-offset reflection seismogram, called a corridor stack, at the hole location. When the seismicVISION data are of very high quality, this stack should compare favorably with the hole location trace in the 3-D multichannel seismic (MCS) reflection volume. Differences in arrival times between these seismograms may indicate problems with the velocity field used to depth migrate/​convert the MCS data.

Depth conversion of seismic reflection data

We used the time-depth transforms obtained from the seismicVISION data to convert the original prestack time-migrated 3-D seismic reflection data to the depth domain for correlation with the LWD data. For each site we made depth conversions of an inline seismic section that passes through the drill site (100 traces on each side of the site using ProMAX seismic processing software).

Synthetic seismograms

We constructed synthetic seismograms using the best available density curve and the detailed slowness (inverse of velocity) log from the sonicVISION tool. Where the sonicVISION tool was not working properly, because of slow formations or other complications, we generated a velocity curve from the density log using the Gardner equation (Gardner et al., 1974).

Conversion from time to depth was required during synthetic seismogram construction to allow correlation of the depth-based LWD logs to the traveltime-based seismic data. This conversion was done using the check shot data generated by the seismicVISION tool to generate a traveltime to depth relationship.

The sonicVISION tool provides good information for interval velocity changes at short wavelengths. The seismicVISION tool, after processing and filtering, provides a reliable smooth interval velocity curve at long wavelengths. Following general industrial practices, we took the smoothed interval velocity curve from the check shot as correct and calibrated the sonic log so that these two curves matched in their time to depth relationships. This calibration is achieved by creating a drift curve that shifts the sonic log within defined intervals (based on inflection points in the drift curve) while preserving the shorter wavelength relative velocity changes provided by the sonic data.

To create a synthetic seismogram, a source wavelet was convolved with a reflectivity series using GeoFrame software (Schlumberger). The reflectivity is expressed in the following form:

where ν₁ , ν₂ and ρ₁ , ρ₂ are the acoustic velocity and density in the upper layer and lower layer, respectively. For Expedition 314, we estimated the source wavelet from the best waveform and amplitude match provided by wavelets extracted within 20 traces of each site in the orientation (inline or cross-line) providing the flattest seafloor. We used a deterministic extraction method based either on the energy or power spectrum and in each case let the software define the best wavelet length and lag. To obtain a wavelet we defined a start and end time within the trace, always starting shallower than the seafloor and ending deeper than the wavelet length. The software computes a zero lag autocorrelation of the reflection coefficients from each trace within the region defined and then computes a signal-to-noise ratio for each trace. An optimal inline/​cross-line pair is returned along with an optimal time lag that results in the highest signal-to-noise ratio and that passes a 90% confidence level based on normalized mean square error (NMSE). We determined the optimal wavelet length by computing the best length at each site using the portion of energy predicted (PEP) method. In this method, a wavelet is generated based on the deterministic method (power spectrum comparison or energy comparison) for each wavelength to be tested and then convolved with the reflection coefficient to generate a synthetic seismogram. The resulting synthetic seismogram is then compared to the input trace based on its PEP and the best match wavelet length is returned. The optimum wavelet lengths in all cases proved to be 256 ms, the lags varied some but were usually close to zero, and the extracted wavelets were all zero phase. For each site, we varied the start and end time for the wavelet extraction window and tried both energy and power spectrum comparisons until we attained a source wavelet that had high signal-to-noise ratio, passed the NMSE test, and visually provided the best fit to the frequency spectrum of the input data.

Using the check shot curve, calibrated sonic log, and best available density log we were able to create a reflection coefficient series. This series was then convolved with the extracted source wavelet to generate a synthetic seismogram at a 4 ms sampling interval. Displaying the synthetic seismogram beside the seismic data from the area of the borehole provides information about specific boundaries of interest and a quality check on velocity and density logs.

Top of page   |    Previous   |    Next